AI Receptionist for a Small Business: How to Set One Up and What to Test Before It Takes Real Calls

Who wrote this: the team behind Voice Ninja, a product built by Wope. We mention it where it fits, alongside other ways to do the same thing.

An AI receptionist for a small business is a voice agent that answers your phone line, handles routine questions, books or changes appointments in your own calendar, and passes anything it cannot handle to a person. To set one up well, decide which calls it covers and when it hands off, connect it to your booking system, and run a set of scripted test calls before it takes real customers.

Switching on an AI receptionist takes an afternoon. Getting one that your regulars do not complain about takes a little more thought. This guide is written by the team behind Voice Ninja, one of Wope's own products, and covers the decisions, the setup steps and the testing we would do for a salon, a dental clinic or a restaurant.

What an AI receptionist can and cannot do

Modern AI phone answering services understand natural speech, cope with callers who change their mind mid-sentence, and can take actions in other systems while the caller is on the line. That makes them useful for most of the calls a front desk takes.

The rule of thumb: if a new hire could handle the call on day one with a printed sheet of answers and access to the booking screen, an AI receptionist can handle it too. If the call needs a manager, it needs a handoff.

Three decisions to make before you build

1. Which calls it handles

List the last 50 calls your front desk took, by reason. For a salon this is usually bookings, rescheduling, "are you open on Sunday?" and "how much is a colour?". For a dental clinic, add new-patient registration and "I have toothache, can I come in today?". For a restaurant, it is reservations, opening hours, dietary questions and large bookings. Start with the three or four reasons that cover most of the volume. Add more later.

2. When it hands off to a person

Write the handoff rules down before you write anything else. Typical triggers:

• The caller asks for a person, twice or once, depending on your preference.
• The caller mentions pain, an emergency, a complaint or a refund.
• The agent has failed to understand the caller twice in a row.
• The request is outside the list of calls it handles.

Then decide what a handoff means at each time of day: a live transfer to the front desk during opening hours, and a detailed message plus a callback promise after hours. A handoff that rings an empty desk is worse than no AI at all.

3. Which system it books into

An AI receptionist that only takes messages saves you some time. One that books directly into your calendar, salon software or reservation system saves far more, but needs a working connection to that system. Check that your booking tool has an API or a supported integration, and decide which fields the agent must collect: name, phone number, service, staff member, date and time, and anything else your team relies on.

Step-by-step setup

1. Write the knowledge sheet. One page: hours, address, parking, services and prices, cancellation policy, what you never do. This is what the agent answers from. Keep it short and factual.
2. Pick a starting template. Most platforms offer templates by industry. Start from the closest one and edit it, rather than writing a script from a blank page.
3. Set the greeting and disclosure. Say the business name and that the caller is speaking to an AI assistant, in one sentence. For example: "Thanks for calling Harbour Dental, this is the clinic's AI assistant. How can I help?"
4. Choose the voice and language. Pick a voice that suits your brand and the languages your callers actually speak. Listen to it on a phone line, not just through laptop speakers.
5. Connect the booking system. Set up the actions the agent can take: check availability, create a booking, move a booking, cancel a booking. Give it read access to what it needs and nothing more.
6. Configure handoff. Add the transfer number for opening hours and the message flow for after hours. Make sure messages arrive somewhere a person checks, such as email or your team chat.
7. Connect the phone number. Either get a new number from the platform or forward your existing number to it. Many businesses start with overflow only: the AI answers when the desk does not pick up within a few rings, or outside opening hours.
8. Test, then go live in stages. Run the checklist below, then start with after-hours calls for a week before taking daytime calls.

The pre-launch test checklist

Write each test as a short script, make the call from a real phone, and record whether the agent got it right. Run the full set again after every change to the prompt, the voice or the integration. Get at least two other people to make calls, ideally with different accents and speaking styles.

• Simple booking. "I'd like a haircut on Thursday afternoon." Check that the booking appears in the calendar with the right service, time and phone number.
• Booking with no availability. Ask for a fully booked slot. It should offer alternatives, not invent one.
• Reschedule and cancel. Move an existing booking, then cancel it. Check it identifies the right booking when the caller has more than one.
• Interruptions. Talk over the agent mid-sentence and change your request. It should stop, listen and follow the new request.
• Spelling and numbers. Give an unusual surname and a phone number with pauses. Check what was saved.
• Accents and noise. Call from a street, a car and a busy kitchen. Use colleagues with different accents.
• Off-script questions. Ask something not on the knowledge sheet. It should say it does not know and offer a message or transfer, not guess.
• Policy pressure. Ask for a discount or a same-day slot that breaks policy. It should not agree to anything it cannot deliver.
• Handoff triggers. Say "I want to speak to a person", mention pain or a complaint. Check the transfer connects during hours and a message is taken after hours.
• After-hours call. Call at night. Check the greeting, the message flow and where the message lands.
• Silence and hang-ups. Say nothing for ten seconds; hang up mid-booking. Check nothing half-finished is left in the calendar.
• Industry specifics. For a restaurant, a party of twelve and a severe allergy. For a clinic, a new patient and an urgent symptom. For a salon, a service that needs a patch test first.

After launch, read a sample of transcripts every week for the first month. The calls that go wrong in real life are rarely the ones you thought to test, and each one becomes a new test script.

Disclosure, recording and consent

Tell callers they are talking to an AI. It is courteous and it reduces complaints. In the EU it is also a legal point: Article 50 of the AI Act requires that people are informed when they interact directly with an AI system, unless that is obvious from the context.

Call recording is the other thing to check. In the US, federal law allows recording with one party's consent, but a number of states require every party to consent, and calls across state lines can be governed by the stricter rule. UK and other European rules differ again. A short line in the greeting such as "calls are recorded for quality" covers most situations, but check the rules where you and your callers are, and ask a lawyer if you work in a regulated field such as healthcare.

What it costs, at a high level

AI receptionist pricing usually has three parts: a per-minute charge for conversation time, a phone number, and any setup or integration work. Some platforms add a monthly subscription; others are pay as you go. Human virtual receptionist services typically charge per call or per minute of a person's time, which costs more per call but brings human judgement.

We have covered per-minute rates, monthly totals for small businesses and the fees that do not appear on pricing pages in a separate guide: AI voice agent cost. The short version is to compare options by cost per call handled correctly, not by the headline rate.

How to compare options

There are three broad categories:

• DIY voice platforms. You build and run the agent yourself, using templates and a visual builder. Lowest running cost and most control; you own the testing.
• Done-for-you AI answering services. A provider sets up and maintains the agent for you. Less work for you, usually a monthly fee.
• Human virtual receptionists. Remote staff answer your calls. Best for complex or sensitive calls; most expensive per call and limited by staff hours.

Questions to ask any provider:

1. Can it book directly into the system we already use, or only take messages?
2. How does handoff work, and what happens if nobody answers the transfer?
3. Can we read transcripts of every call?
4. Which languages and voices are available, and can we hear them on a real call?
5. Is there a monthly minimum or contract, and what happens in our busiest month?
6. How do we test changes before they reach real callers?

Where Voice Ninja fits

Voice Ninja is our DIY voice platform, so it sits in the first category above. It builds AI phone agents for inbound and outbound calls and voice widgets for websites. It includes templates for clinics, hotels and restaurants, a choice of language model, language and voice for each agent, and custom tool calls so an agent can book, look up and update records in your own systems through your APIs. Agents can transfer a caller to another agent without losing context, and analytics show captured leads and per-agent performance, with transcripts. Pricing is credit-based: you pay for what you use, with no monthly commitment. Developers get API and WebSocket access. It is trusted by 5k+ users.

It is a good fit if you want to set up and test the receptionist yourself. If you would rather someone else built, connected and tested it, a done-for-you service or a human receptionist may suit you better. More detail is on the Voice Ninja product page, and if you are weighing up other uses of AI across the business, see AI agents for small business.

Frequently asked questions

What can an AI receptionist do for a small business?

It can answer every call, handle routine questions about hours, prices and location, book, move or cancel appointments in your calendar or booking system, take messages and transfer callers to a person. It should not give medical, legal or financial advice or make promises your staff have not approved.

How long does it take to set up an AI receptionist?

A basic receptionist that answers questions and takes messages can be live in a day using a template. One that books into your real calendar and hands off cleanly usually takes one to three weeks, most of it spent on the booking integration and testing.

Should an AI receptionist tell callers it is an AI?

Yes. It is good practice everywhere, and in the EU the AI Act requires people to be informed when they are interacting with an AI system unless it is obvious. Call recording consent rules also vary by country and US state, so check the rules that apply to you.

How much does an AI receptionist cost?

Most AI receptionist platforms charge per minute of conversation, plus a phone number and any setup or integration work. For a small business taking a few hundred to a couple of thousand minutes a month, see our AI voice agent cost guide for a full breakdown.

What is the difference between an AI answering service and a virtual receptionist?

An AI answering service uses software to answer and act on calls at any hour. A virtual receptionist service uses remote human staff. Many businesses combine them: the AI handles routine calls and passes anything unusual to a person.